Automatic scoring of drug-induced sleep endoscopy for obstructive sleep apnea using deep learning.

BACKGROUND: Treatment of obstructive sleep apnea is crucial for long term health and reduced economic burden. For those considered for surgery, drug-induced sleep endoscopy (DISE) is a method to characterize location and pattern of sleep-related upper airway collapse. According to the VOTE classification system, four upper airway sites of collapse are characterized: velum (V), oropharynx (O), tongue (T), and epiglottis (E). The degree of obstruction per site is classified as 0 (no obstruction), 1 (partial obstruction), or 2 (complete obstruction). Here we propose a deep learning approach for automatic scoring of VOTE obstruction degrees from DISE videos. METHODS: We included 281 DISE videos with varying durations (6 s-16 min) from two sleep clinics: Copenhagen University Hospital and Stanford University Hospital. Examinations were split into 5-s clips, each receiving annotations of 0, 1, 2, or X (site not visible) for each site (V, O, T, and E), which was used to train a deep learning model. Predicted VOTE obstruction degrees per examination was obtained by taking the highest predicted degree per site across 5-s clips, which was evaluated against VOTE degrees annotated by surgeons. RESULTS: Mean F1 score of 70% was obtained across all DISE examinations (V: 85%, O: 72%, T: 57%, E: 65%). For each site, sensitivity was highest for degree 2 and lowest for degree 0. No bias in performance was observed between videos from different clinicians/hospitals. CONCLUSIONS: This study demonstrates that automating scoring of DISE examinations show high validity and feasibility in degree of upper airway collapse.

[Narcolepsy: From the discovery of a wake promoting peptide to autoimmune T cell biology and molecular mimicry with flu epitopes]. / Narcolepsie : une maladie auto-immune affectant un peptide de l'éveil liée à un mimétisme moléculaire avec des épitopes du virus de la grippe.

Narcolepsy-cataplexy was first described in the late 19th century in Germany and France. Prevalence was established to be 0.05 % and a canine model was discovered in the 1970s. In 1983, a Japanese study found that all patients carried HLA-DR2, suggesting autoimmunity as the cause of the disease. Studies in the canine model established that dopaminergic stimulation underlies anti-narcoleptic action of psychostimulants, while antidepressants were found to suppress cataplexy through adrenergic reuptake inhibition. No HLA association was found in canines. A linkage study initiated in 1988 revealed in hypocretin (orexin) receptor two mutations as the cause of canine narcolepsy in 1999. In 1992, studies on African Americans showed that DQ0602 was a better marker than DR2 across all ethnic groups. In 2000, hypocretin-1/orexin A levels were measured in the cerebrospinal fluid (CSF) and found to be undetectable in most patients, establishing hypocretin deficiency as the cause of narcolepsy. Decreased CSF hypocretin-1 was then found to be secondary to the loss of the 70,000 neurons producing hypocretin in the hypothalamus, suggesting immune destruction of these cells as the cause of the disease. Additional genetic studies, notably genome wide associations (GWAS), found multiple genetic predisposing factors for narcolepsy. These were almost all involved in other autoimmune diseases, although a strong and unique association with T cell receptor (TCR) alpha and beta loci were observed. Nonetheless, all attempts to demonstrate presence of autoantibodies against hypocretin cells in narcolepsy failed, and the presumed autoimmune cause remained unproven. In 2009, association with strep throat infections were found, and narcolepsy onsets were found to occur more frequently in spring and summer, suggesting upper away infections as triggers. Following reports that narcolepsy cases were triggered by vaccinations and infections against influenza A 2009 pH1N1, a new pandemic strain that erupted in 2009, molecular mimicry with influenza A virus was suggested in 2010. This hypothesis was later confirmed by peptide screening showing higher activity of CD4+ T cell reactivity to a specific post-translationally amidated segment of hypocretin (HCRT-NH2) and cross-reactivity of specific TCRs with a pH1N1-specific segment of hemagglutinin that shares homology with HCRT-NH2. Strikingly, the most frequent TCR recognizing these antigens was found to carry sequences containing TRAJ24 or TRVB4-2, segments modulated by narcolepsy-associated genetic polymorphisms. Cross-reactive CD4+ T cells with these cross-reactive TCRs likely subsequently recruit CD8+ T cells that are then involved in hypocretin cell destruction. Additional flu mimics are also likely to be discovered since narcolepsy existed prior to 2009. The work that has been conducted over the years on narcolepsy offers a unique perspective on the conduct of research on the etiopathogeny of a specific disease.

TITLE: Narcolepsie : une maladie auto-immune affectant un peptide de l'éveil liée à un mimétisme moléculaire avec des épitopes du virus de la grippe. ABSTRACT: La narcolepsie et la cataplexie sont décrites pour la première fois à la fin du XIXe siècle en Allemagne et en France. La prévalence de la maladie est établie à 0,05 % et un modèle canin est découvert dans les années 1970. En 1983, une étude japonaise révèle que les patients narcoleptiques sont porteurs d'un marqueur génétique unique, l'antigène leucocytaire HLA-DR2, suggérant l'auto-immunité comme cause de la maladie. Il faudra attendre 1992 pour qu'il soit montré, grâce à une étude chez des patients afro-américains, que DQ0602, un autre gène HLA, est la véritable cause de cette association. Des études pharmacologiques conduites sur le modèle canin établissent que la stimulation dopaminergique est le mode d'action des stimulants sur l'éveil, tandis que les antidépresseurs suppriment la cataplexie en inhibant la recapture adrénergique. Aucune association HLA n'est cependant mise en évidence chez les chiens, suggérant une cause distincte de la maladie humaine. Une étude de liaison génétique chez les chiens, initiée en 1988, révèle en 1999 que la narcolepsie canine est causée par des mutations du récepteur 2 de l'hypocrétine (orexine). En 2000, l'hypocrétine-1/orexine A est mesurée dans le liquide céphalo-rachidien (LCR) et on découvre qu'elle est indétectable chez la plupart des patients narcoleptiques, établissant qu'un déficit hypocrétinergique est la cause de la narcolepsie humaine. La diminution de l'hypocrétine-1 dans le LCR, secondaire à la perte des 70 000 neurones hypothalamiques produisant l'hypocrétine, est démontrée, ce qui, avec l'association au locus HLA, suggère qu'une destruction immunitaire de ces cellules est la cause de la maladie. D'autres études génétiques, notamment d'association à l'échelle du génome (GWAS), révèlent l'existence de nombreux facteurs génétiques prédisposant à la narcolepsie, la plupart étant également impliqués dans d'autres maladies auto-immunes. Une association forte et unique avec les loci des récepteurs lymphocytaires T (TCR) alpha et bêta est aussi observée, suggérant un rôle prépondérant des lymphocytes T. En dépit de nombreux efforts, toutes les tentatives visant à démontrer la présence d'auto-anticorps contre les cellules à hypocrétine dans la narcolepsie échouent, et la cause auto-immune présumée de cette maladie reste à l'état d'hypothèse. À la suite de la grippe pandémique influenza A pH1N1 en 2009, de nombreux cas de narcolepsie apparaissent, suggérant un mimétisme moléculaire avec le virus de la grippe qui pourrait déclencher la maladie auto-immune. Cette hypothèse est confirmée par un criblage peptidique montrant une plus grande réactivité des lymphocytes T CD4+ à un segment spécifique de l'hypocrétine (HCRTNH2) et une réactivité croisée des TCR correspondants à un segment d'hémagglutinine de pH1N1 qui partage une homologie avec HCRTNH2. De façon remarquable, le TCR le plus fréquent dans la population et qui reconnaît ces antigènes contient des séquences TRAJ24 ou TRVB4-2, segments modulés par des polymorphismes génétiques associés à la narcolepsie dans les études GWAS. Il est probable que les lymphocytes T CD4+ autoréactifs avec HCRTNH2 recrutent par la suite des lymphocytes T CD8+ qui détruisent les cellules à hypocrétine. On peut s'attendre à ce que d'autres séquences mimiques grippales inconnues soient découvertes prochainement puisque la narcolepsie existait avant 2009. Ces découvertes démontrent enfin la cause auto-immune de la narcolepsie. Les travaux menés au cours des années sur la narcolepsie offrent une perspective unique sur la conduite de la recherche sur l'étiopathogénie d'une maladie bien identifiée.

Non-invasive machine learning estimation of effort differentiates sleep-disordered breathing pathology.

OBJECTIVE: Obstructive sleep-disordered breathing (SDB) events, unlike central events, are associated with increased respiratory effort. Esophageal pressure (P es) monitoring is the gold standard for measuring respiratory effort, but it is typically poorly tolerated because of its invasive nature. The objective was to investigate whether machine learning can be applied to routinely collected non-invasive, polysomnography (PSG) measures to accurately model peak negative P es. APPROACH: One thousand one hundred and nineteen patients from the Stanford Sleep Clinic with PSGs containing P es served as the sample. The selected non-invasive PSG signals included nasal pressure, oral airflow, thoracoabdominal effort, and snoring. A long short-term memory neural network was implemented to achieve a context-based mapping between the non-invasive features and the P es values. A hold-out dataset served as a prospective validation of the algorithm without needing to undertake a costly new study with the impractically invasive P es. MAIN RESULTS: The median difference between the measured and predicted P es was 0.61 cmH2O with an interquartile range (IQR) of 2.99 cmH2O and 5th and 95th percentiles of -5.85 cmH2O and 5.47 cmH2O, respectively. The model performed well when compared to actual esophageal pressure signal (ρ median = 0.581, p  = 0.01; IQR = 0.298; ρ 5% = 0.106; ρ 95% = 0.843). SIGNIFICANCE: A significant difference in predicted P es was shown between normal breathing and all obstructive SDB events; whereas, central apneas did not significantly differ from normal breathing. The developed system may be used as a tool for quantifying respiratory effort from the existing clinical practice of PSG without the need for P es, improving characterization of SDB events as obstructive or not.

History of narcolepsy at Stanford University.

Although narcolepsy was first described in the late nineteenth century in Germany and France, much of the research on this disorder has been conducted at Stanford University, starting with Drs. William C. Dement and Christian Guilleminault in the 1970s. The prevalence of narcolepsy was established, and a canine model discovered. Following the finding in Japan that almost all patients with narcolepsy carry a specific HLA subtype, HLA-DR2, Hugh Mac Devitt, F. Carl Grumet, and Larry Steinman initiated immunological studies, but results were generally negative. Using the narcoleptic canines, Dr. Nishino and I established that stimulants increased wakefulness by stimulating dopaminergic transmission while antidepressants suppress cataplexy via adrenergic reuptake inhibition. A linkage study was initiated with Dr. Grumet in 1988, and after 10 years of work, the canine narcolepsy gene was cloned by in 1999 and identified as the hypocretin (orexin) receptor 2. In 1992, studying African Americans, we also found that DQ0602 rather than DR2 was a better marker for narcolepsy across all ethnic groups. In 2000, Dr. Nishino and I, in collaboration with Dr. Lammers in the Netherlands, found that hypocretin 1 levels in the cerebrospinal fluid (CSF) were undetectable in most cases, establishing hypocretin deficiency as the cause of narcolepsy. Pursuing this research, our and Dr. Siegel's group, examining postmortem brains, found that the decreased CSF hypocretin 1 was secondary to the loss the 70,000 neurons producing hypocretin in the hypothalamus. This finding revived the autoimmune hypothesis but attempts at demonstrating immune targeting of hypocretin cells failed until 2013. At this date, Dr. Elisabeth Mellins and I discovered that narcolepsy is characterized by the presence of autoreactive CD4(+) T cells to hypocretin fragments when presented by DQ0602. Following reports that narcolepsy cases were triggered by vaccinations and infections against influenza A 2009 pH1N1, a new pandemic strain that erupted in 2009, our groups also established that a small epitope of pH1N1 resembles hypocretin and is likely involved in molecular mimicry. Although much remains to be done, these achievements, establishing hypocretin deficiency as the cause of narcolepsy, demonstrating its autoimmune basis, and showing molecular mimicry between hypocretin and sequences derived from a pandemic strain of influenza, are likely to remain classics in human immunology.

A practical guide to the therapy of narcolepsy and hypersomnia syndromes.

Narcolepsy and other syndromes associated with excessive daytime sleepiness can be challenging to treat. New classifications now distinguish narcolepsy/hypocretin deficiency (also called type 1 narcolepsy), a lifelong disorder with well-established diagnostic procedures and etiology, from other syndromes with hypersomnolence of unknown causes. Klein-Levin Syndrome, a periodic hypersomnia associated with cognitive and behavioral abnormalities, is also considered a separate entity with separate therapeutic protocols. Non hypocretin-related hypersomnia syndromes are diagnoses of exclusion. These diagnoses are only made after eliminating sleep deprivation, sleep apnea, disturbed nocturnal sleep, and psychiatric comorbidities as the primary cause of daytime sleepiness. The treatment of narcolepsy/hypocretin deficiency is well-codified, and involves pharmacotherapies using sodium oxybate, stimulants, and/or antidepressants, plus behavioral modifications. These therapies are almost always needed, and the risk-to-benefit ratio is clear, notably in children. Detailed knowledge of the pharmacological profile of each compound is needed to optimize use. Treatment for other syndromes with hypersomnolence is more challenging and less codified. Preferably, therapy should be conservative (such as modafinil, atomoxetine, behavioral modifications), but it may have to be more aggressive (high-dose stimulants, sodium oxybate, etc.) on a case-by-case, empirical trial basis. As cause and evolution are unknown in these conditions, it is important to challenge diagnosis and therapy over time, keeping in mind the possibility of tolerance and the development of stimulant addiction. Kleine-Levin Syndrome is usually best left untreated, although lithium can be considered in severe cases with frequent episodes. Guidelines are provided based on the literature and personal experience of the author.